Service status

Daily availability and first-token time of the RoamAI API, updated every 10 minutes. Counting starts on 17 October 2026 and then covers the last 30 days.

Availability since 2026-10-17

99.03%

First-token time today (median)

2.8 s

Availability by day

2026-10-172026-10-30
  • 99% or more
  • 95% to 99%
  • below 95%
  • no requests that day

Day by day

  1. 10-30

    Availability
    99.86%
    First token
    2.8 s

  2. 10-29

    Availability
    97.73%
    First token
    6.2 s

  3. 10-28

    Availability
    99.15%
    First token
    8.7 s

  4. 10-27

    Availability
    99.53%
    First token
    7.8 s

  5. 10-26

    Availability
    98.28%
    First token
    8.7 s

  6. 10-25

    Availability
    95.90%
    First token
    7.5 s

  7. 10-24

    Availability
    100%
    First token
    9.6 s

  8. 10-23

    Availability
    99.93%
    First token
    8.8 s

  9. 10-22

    Availability
    99.92%
    First token
    8.1 s

  10. 10-21

    Availability
    96.37%
    First token
    9.5 s

  11. 10-20

    Availability
    100%
    First token
    5.7 s

  12. 10-19

    Availability
    100%
    First token
    6.3 s

  13. 10-18

    Availability
    –
    First token
    –

  14. 10-17

    Availability
    100%
    First token
    23.4 s

Updated 2026-10-30 14:57 (Beijing time)

How the numbers are counted

  • Availability = successful requests ÷ (successful requests + requests that failed on our side + requests turned away because we were at capacity).
  • Requests that were invalid in themselves, such as a bad parameter or an unknown model, and requests the caller cancelled are not counted. They say nothing about the service.
  • A request that hit an error and then succeeded on an automatic retry counts as a success.
  • First-token time is the time from receiving a request to returning its first token, as the median of all successful requests that day. It includes the time the model spends thinking.
  • Days run from midnight to midnight Beijing time (UTC+8).